Third Trimester Cortisol is Positively Associated with Gestational Weight Gain in Pregnant Women with Class One Obesity

Abstract

Background/Objective:

Prevalence of pre-pregnancy obesity and excessive gestational weight gain (GWG) are higher among women of color with low SES. Dysregulation of the Hypothalamic-Pituitary-Adrenal (HPA) axis and its end-product, cortisol, during pregnancy is hypothesized to be associated with excessive GWG. However, past studies have produced inconsistent findings and often did not include health disparities populations. This study examined the association between pre-pregnancy body mass index (BMI), third trimester diurnal cortisol, and GWG in low-income, predominantly Hispanic women.

Subjects/Methods:

The MADRES study is an ongoing prospective cohort study of primarily Hispanic, low-income pregnant women and their children in Los Angeles, California. Data from 176 participants were included in this study. Total cortisol secretion (area under the curve, AUC) was quantified using four salivary cortisol samples (awakening, 30 minutes after awakening, afternoon, and bedtime) that were collected at home on one day during the third trimester of pregnancy. Moderation of the association between total cortisol and GWG by pre-pregnancy BMI was tested using multiple linear regression with a multiplicative interaction term.

Results:

There was no association between total cortisol secretion and GWG overall (p=0.82), but the association between total cortisol and GWG was stronger for women with class 1 pre-pregnancy obesity compared to women with normal pre-pregnancy BMI (interaction term p=0.04).

Conclusions:

Results suggest that obesity status before pregnancy may be exacerbating the physiological impact of cortisol on GWG.

Introduction

Gestational weight gain (GWG) plays an important role in a healthy pregnancy. The Institute of Medicine (IOM) has recommended GWG guidelines based on a mother’s pre-pregnancy body mass index (BMI; weight [kg]/ height [m]²) to promote optimal prenatal health outcomes. Currently in the United States, almost half the women of childbearing age are overweight or obese. Mothers who are overweight and obese prior to pregnancy are more likely to gain weight beyond the recommended guidelines (i.e., excessive GWG).¹ While about one in five pregnant women gain inadequate gestational weight, almost half gain excessive weight during pregnancy.² Excessive GWG increase the risk of pregnancy complications, adverse birth outcomes, and long-term chronic health conditions for both the mother and child.^3–5 Given these alarmingly high rates and downstream risks of excessive GWG, elucidating physiological correlates of excessive GWG is of great interest in improving prenatal health outcomes.

A growing body of literature suggests dysregulation of the Hypothalamic-Pituitary-Adrenal (HPA) axis during pregnancy is associated with pregnancy outcomes, including excessive GWG.^6,7 The HPA axis is a major component of the neuroendocrine system that comprises a synergistic cascade of enzymes and hormones that ultimately stimulates the adrenal glands to secrete cortisol.^8,9 Cortisol is one of the key hormones in the “fight or flight” response that is activated when the HPA axis is triggered under acute stress – both psychological and physiological—and homeostatic balance is disrupted.¹⁰ In addition to its rapid peaks in reaction to stressors, cortisol levels follow a circadian pattern, characterized by increasing levels at awakening, followed by a steep peak at 30–45 minutes after awakening, and a slow decline until bedtime.^11,12 Researchers often measure this diurnal rhythm through repeated sampling of saliva throughout the day, as saliva collection is non-invasive and easily self-administered outside the laboratory setting.¹³ Cortisol levels are then examined at each assessment point or integrated across the day using the area under the curve respective to ground (AUCg), which captures the circadian changes in cortisol and overall secretion throughout the day.^14,15 In non-pregnant populations, there is substantial evidence supporting the relationship between dysregulated circadian cortisol patterns, obesity, and weight gain.⁹ Individuals with abdominal obesity show both overactivation of the HPA axis and reduced diurnal variation in cortisol levels. ¹⁶ In addition, cortisol is causally linked to the accumulation of fat cells and increased appetite, consequently contributing to weight gain and the development of obesity.¹⁷ Therefore, it is possible that pre-pregnancy BMI, diurnal cortisol, and GWG are similarly related during the pregnancy period.

However, there is a dearth of literature on this topic that takes into account the unique health implications of excessive weight gain and changes in HPA axis functioning during pregnancy. For example, though cortisol circadian pattern remains unchanged, maternal cortisol levels increase two to four-fold during pregnancy due to additional cortisol production from the maternal HPA axis, in conjunction with the activation of the fetal HPA axis, and placental production of corticotropin-releasing hormone (CRH).^18–20 This increase in cortisol secretion supports the maturation of fetal organs, such as the lungs and liver, and preparation for the increase in fetal metabolic rate and thermogenesis that are crucial for the transition to life outside the womb.²¹ Furthermore, glucocorticoids, including cortisol, play a pivotal role in the initiation and maintenance of parturition.²² These unique alterations and implications of HPA axis functioning during the prenatal period make pregnant women an important population to examine obesity, HPA axis functioning, and weight gain, separately from non-pregnant populations.

To date, there have only been a handful of studies on this topic, and they have produced inconsistent findings. While some found higher pre-pregnancy BMI to be associated with lower cortisol levels earlier in the day (i.e. awakening and early afternoon),²³ others only identified higher evening cortisol levels in participants with obesity compared to those without.²⁴ In addition, some have found no association between cortisol and GWG, regardless of pre-pregnancy obesity status,²⁵ while others report increased evening cortisol levels in pregnant women with obesity who gain excessive gestational weight.²⁴ It must also be noted that these existing studies have been mostly conducted in Non-Hispanic White or Caucasian women and women of high socioeconomic status, even though there are known disparities in rates of overweight/obesity and excessive GWG in the United States. Approximately 39.8% of Non-Hispanic White women of childbearing age have overweight or obesity, while these rates are as high as 43.7% in Hispanic women and 56.9% in Black women.²⁶ And although the average prevalence of excessive GWG in Non-Hispanic White women is 35%, studies report prevalence as high as 53% in low-income women, 51% in Hispanic women, and 61% in Black women.^27,28 Therefore, understanding the correlates of excessive GWG among minority and low-income women is crucial in reducing the unequitable burden of disease they bear. Thus, due to the small number of studies, inconsistent findings, and lack of research in diverse populations, the psychobiological underpinnings of pre-pregnancy obesity status, diurnal cortisol, and GWG remain elusive in those who bear the unequitable burden of disease—minority and low-income women.

This study examined the association between pre-pregnancy BMI, third trimester total cortisol secretion, and GWG in the Maternal and Developmental Risks from Environmental and Social Stressors (MADRES) pregnancy cohort study.²⁹ This study’s main aim was to investigate the association between the total cortisol secretion (AUCg) and GWG, while testing for effect modification of this relationship by pre-pregnancy BMI. We hypothesized that there would be a positive association between third trimester AUCg and GWG, and that this relationship would be stronger in those with higher pre-pregnancy BMI. We also conducted an ancillary aim to examine differences in AUCg and diurnal cortisol levels (i.e., cortisol levels at each collection time point throughout the day) by pre-pregnancy BMI status. We hypothesized that AUCg levels would be more elevated in those of higher pre-pregnancy BMI categories.

Methods

Sample

The MADRES study is an ongoing prospective cohort study of primarily Hispanic, low-income pregnant women and their children in Los Angeles, California. Details of the MADRES protocol have been described elsewhere.²⁹ Participants were recruited from Los Angeles County + USC (LAC+USC) Medical Center, the Women’s Health Center at Eisner Health, and the South-Central Family Health Center. Inclusion criteria were: (1) <30 weeks pregnant at the time of enrollment, (2) ≥18 years of age, (3) singleton pregnancy, and (4) English or Spanish speaking. Exclusion criteria were (1) HIV positive status; (2) physical, mental, or cognitive disabilities that prevent participation; (3) current incarceration; or (4) multiple gestation. Maternal consent and HIPAA authorization for abstracting electronic medical records (EMR) was obtained prior to any study assessment. The Institutional Review Board at the University of Southern California approved all aspects of this study.

For women enrolling at less than 20 weeks gestation (“regular entry”), the prenatal data collection protocol consisted of 1) initial in-person visit within 2 weeks of recruitment; 2) medical records abstraction from prenatal care clinic visits; 3) telephone interviews between 18–27 weeks gestation; and 4) a third trimester in-person visit between 30–34 weeks gestation. Both in-person visits consisted of an interviewer-administered questionnaire in English or Spanish and anthropometric measurements of height and weight. The third trimester in-person visit additionally included collection of the saliva samples from the participants for cortisol measurement. For women enrolling between 20–30 weeks gestation (“late entry”), the prenatal in-person visit, which consisted of the interviewer-administered questionnaire, anthropometric measures of height and weight, and saliva sample collection, occurred between 28–36 weeks. Furthermore, some of the women also participated in a sub-study (i.e. the real-time personal monitoring sub study) where additional personal monitoring data and repeated saliva sampling were collected during the first/second trimester and third trimester.³⁰

Measures

Salivary Cortisol

Women collected saliva samples on one day during the third trimester (i.e., 24 weeks to 35 weeks gestation) in their natural environment, as they went about their usual routines. Participants used a Salivette device (Sarstedtf, Inc. Rommelsdolf, Germany), which is a small cotton dental roll that participants gently chew for two minutes to collect saliva. Most participants (n=153, 86.9%) collected saliva samples on a weekday. A total of four saliva samples were collected to capture the diurnal pattern of cortisol secretion during one at each of the following timepoints: awakening, 30 min after awakening, afternoon (around 3:00pm), and bedtime. ³¹ Participants were informed that compliance to the morning sampling protocol is important for quantifying cortisol.³² Of the 176 women, 8 (4.7%) waited more than 15 minutes after awakening to collect the first sample, and 12 (6.8%) waited less than 15 minutes or more than 45 minutes after the first sample to collect the second sample. At the time of collection, participants directly noted on the saliva tube the date, time, and whether any eating, drinking (besides water), tooth brushing, smoking or exercising had occurred in the prior 30 min. The contamination rate was lowest for awakening samples (n=5, 2.9%) and most common for afternoon (3:00pm) samples (n=66, 37.9%). Samples were flagged if participants had waited more than 15 minutes after awakening to collect their first sample, waited less than 15 minutes or more than 45 minutes after the first sample to collect their second sample, and if dietary eating, drinking (besides water), tooth brushing, smoking or exercising occurred in the prior 30 minutes. Participants stored their samples in their refrigerators for an average of 6 days (range: 3 – 47 days) until their third trimester in-person visits when the samples were transferred to a laboratory freezer for storage at −80°C. The number of days that the samples were stored in the participants’ home refrigerators did not affect salivary cortisol levels, r (175) =−0.12, p=0.11. Samples were sent to a commercial laboratory in batches and were assayed with chemiluminescence immunoassay (CLIA; IBL International, Hamburg, Germany), which has a lower detection limit of .005 ug/dL and intra- and inter-assay coefficients in the range of 3.0 – 4.1%.

We calculated the AUCg using the standard trapezoidal formula: ($AUC={\sum }_{i=1}^4\frac{\left(m_{\left(i+1\right)}+m_i\right)\times t_i}{2}$) with t_i denoting the time difference between measurement i and i+11, and m_i denoting cortisol at measurement i.¹¹ The AUCg allows us to assess the overall secretion of cortisol on the day by incorporating the four repeated measurements. For the ancillary analysis, diurnal cortisol was quantified by examining cortisol levels individually at each of the four saliva collection timepoints throughout the day (i.e., awakening, 30 min after awakening, at afternoon (around 3:00pm), and bedtime). Non-adherence to the morning cortisol protocol nor contamination did not affect AUCg or diurnal cortisol levels.

Maternal Height and Weight

Maternal weight and height during pregnancy were abstracted from EMRs and measured by trained staff during each of the prenatal in-person study visits using an electronically-calibrated digital scale (Tanita, Perspective Enterprises, Portage, MI) and a commercial stadiometer (Model PE-AIM-101, Perspective Enterprises) to the nearest 0.1 kg and 0.1 cm, respectively.

Total GWG was defined as the difference between a mother’s weight measured within two weeks before giving birth and her pre-pregnancy weight. Self-reported pre-pregnancy weight was ascertained through interviewer-administered questionnaires during pregnancy. If missing, then the first weight of the index pregnancy (obtained from maternal EMR) was used in lieu of self-reported pre-pregnancy weight. Participants were flagged if there was a +/− 10kg difference in self-reported weight versus first measured weight or if the first weight was obtained > 10 weeks gestational age. Self-reported pre-pregnancy weight and height were used to calculate the pre-pregnancy BMI (kg/m2) and classified using CDC categories: underweight (BMI< 18.5), normal weight (BMI ≥18.5 and < 25), overweight (BMI ≥25 and < 30), class 1 obese (BMI ≥30 and < 35), and class 2/3 obese (BMI ≥35).³³ We excluded women whose pre-pregnancy BMI was categorized as underweight (n=5), and women with class 2 and class 3 obesity were combined, given the small sample size in each category and similar GWG patterns.

Covariates

We identified covariates a priori based on existing literature and then conducted preliminary analyses to examine their unadjusted associations with GWG and AUCg to determine the final analytic model. The final list of covariates included study design, demographic, and pregnancy-related variables as covariates in this analysis. Study design variables included recruitment site, whether the participant was part of the real-time personal monitoring sub-study, and whether the participant entered the study at ≥20 weeks gestation (i.e. “late entry” vs “regular entry”). Demographic covariates were mother’s ethnicity by birthplace, earliest ascertained education level, and age at baseline. These demographic characteristics were assessed via interviewer-administered questionnaire in English or Spanish during the initial study visit. Pregnancy-related variables included gestational age at delivery, which was recorded in the mother’s EMR, and nausea or vomiting during the third trimester. Mother’s nausea was measured using the Pregnancy-Unique Quantification of Emesis (PUQE) scoring system, which quantifies the severity of nausea or vomiting during the previous 24 hours. ³⁴ The PUQE score was dichotomized to 0=no nausea or vomiting and 1= mild to moderate nausea or vomiting. It should be noted that the PUQE was administered at both the initial visit (if enrolled prior to 20 weeks) and at the third trimester study visit by questionnaire. However, we only used the data from the third trimester study visit in order to increase sample size (as women who enrolled between 20–30 weeks only had data from the third trimester), and third trimester nausea and vomiting were more strongly correlated with third trimester cortisol and GWG (data not shown). Other variables, such as parity, income, gestational age at cortisol sampling, gestational diabetes, hypertension, and physical activity were considered, but based on preliminary analyses showing no evidence of confounding, collinearity issues with other covariates, and model parsimony, we did not include these variables in our final model.

Statistical Analysis

All data analyses were conducted using SAS v9.4.³⁵ Initial descriptive analyses were conducted to examine the distributions of all variables and multicollinearity between exposures. Additional analyses of residual distributions determined whether modeling assumptions were met. Given that the homoscedasticity assumption was met, we proceeded without transforming any variables of interest.³⁶ We conducted chi-square tests, student’s t-tests, and correlations to examine whether any demographic variables were associated with general data missingness or noncompliance with saliva collection protocol. For the ancillary aim, we conducted analysis of variance (ANOVA) and Tukey’s Test to examine differences in AUC and individual cortisol levels at each sampling time by pre-pregnancy BMI status. For the main aim, we used two models to examine the relationship between cortisol AUC and GWG. In model 1, we examined the overall association between third trimester AUCg and GWG using multiple linear regression adjusting for pre-pregnancy BMI, real-time personal monitoring sub-study participation, gestational age at study entry, recruitment site, gestational age at birth, ethnicity by birthplace, highest education level, age, and third trimester nausea or vomiting. In model 2, we included a multiplicative interaction term between AUCg and pre-pregnancy BMI (reference: normal BMI) to examine potential differences in the association between third trimester AUCg and GWG by pre-pregnancy BMI. Model 2 also controlled for the same covariates as model 1. Lastly, we conducted a stratified analysis examining the association between third trimester AUCg and total GWG in each pre-pregnancy BMI category; this further tested the AUCg and total GWG relationship while parsing out any remaining effect of pre-pregnancy BMI. All statistical significance was examined using two-sided tests with α=0.05.

Results

Sample

For this analysis, 433 women had complete GWG data (i.e. women who had already given birth, and their final gestational weight was measured within two weeks of birth). We excluded any women whose self-reported pre-pregnancy weight was more than 10kg discrepant from the first weight during pregnancy (measured by MADRES staff or abstracted from the EMR) (n=19). All 19 women had self-reported pre-pregnancy weights that were lower than the first weight during pregnancy. We also excluded any women who had not completed the third trimester saliva collection (n=173) or failed to provide us with all four saliva samples needed to calculate total cortisol secretion for that day (n=60). We also did not include data from the five women whose self-reported pre-pregnancy BMI was categorized as underweight, as the small sample size, in comparison to the women in other BMI categories, prevented any meaningful interpretation of their results in the final model. Therefore, the final analytical sample consisted of 176 women. The consort diagram illustrating sample exclusion can be found in Figure 1.

Figure 1.

Consort Diagram of Included Observations

Step by step data creation process illustrating included and excluded observations for final data set

Note. Trendline was plotted using locally weighted scatterplot smoothing (LOESS) for each pre-pregnancy BMI category

Descriptive Characteristics

The majority of women identified as US-born Hispanic (n=77, 43.8%) or foreign-born Hispanic (n=71, 40.3%), and about one in four women (n=44, 25.0%) did not finish high school. For about one in three women, this was their first pregnancy (n=56, 33.7%). The average age of the participants was 28.6 years (SD=6.0), and average gestational age at delivery was 39.0 weeks (SD=1.6). One in four women (n=44, 25.0%) had a normal pre-pregnancy BMI, 36.4% (n=64) were categorized as overweight BMI, 25.0% (n=44) were categorized as class 1 obese, and 13.6% (n=24) were categorized as class 2 or 3 obese. Less than half of the women (n=66, 37.5%) reported nausea and/or vomiting during the third trimester. Descriptive characteristics by pre-pregnancy BMI category can be found in Table 1. Furthermore, these characteristics did not significantly differ between the participants with complete GWG data and subset of participants in this current study (Supplementary Table 1).

Table 1.

Descriptive Characteristics of Participants by Pre-Pregnancy Body Mass Index (BMI) Category

	All Participants in Study	Normal BMI BMI ≥18.5 and < 25	Overweight BMI BMI ≥25 and < 30	Obese Class 1 BMI BMI ≥30 and < 35	Obese Class 2/3 BMI BMI ≥35
Sample Size	n=176	n=44 (25.0%)	n=64 (36.4%)	n=44 (25.0%)	n=24 (13.6%)
Variable	n (%) or mean (SD)	n (row %) or mean (SD)	n (row %) or mean (SD)	n (row %) or mean (SD)	n (row %) or mean (SD)
Ethnicity by nativity, n(%)
Non-Hispanic	18 (10.2%)	6 (33.3%)	5 (27.8%)	2 (11.1%)	5 (27.8%)
US-Born Hispanic	77 (43.8%)	20 (26.0%)	21 (27.3%)	24 (31.2%)	12 (15.6%)
Foreign-Born Hispanic	71 (40.3%)	16 (22.5%)	34 (47.9%)	15 (21.1%)	6 (8.5%)
Missing	10 (5.7%)
Highest education level, n(%)
Less than 12th grade	44 (25.0%)	11 (25.0%)	15 (34.1%)	16 (36.4%)	2 (4.6%)
High school	59 (33.5%)	15 (25.4%)	21 (35.6%)	10 (17.0%)	13 (22.0%)
Some college or technical school	44 (25.0%)	12 (27.3%)	14 (31.8%)	13 (29.6%)	5 (11.4%)
Completed 4 years of college	22 (12.5%)	3 (13.6%)	12 (54.6%)	4 (18.2%)	3 (13.6%)
Some graduate training	3 (1.7%)	2 (66.7%)	0 (0.0%)	0 (0.0%)	1 (33.3%)
Missing	4 (2.3%)
Parity (i.e. number of previous births)
First	56 (33.7%)	14 (25.05)	20 (35.7%)	15 (26.8%)	7 (12.5%)
Second	50 (30.1%)	16 (32.0%)	14 (28.0%)	12 (24.0%)	8 (16.0%)
Third	30 (18.1%)	6 (20.0%)	13 (43.3%)	5 (16.7%)	6 (20.0%)
Fourth	20 (12.1%)	4 (20.0%)	7 (35.0%)	8 (40.0%)	1 (5.0%)
Fifth	8 (4.2%)	2 (28.6%)	4 (57.1%)	0 (0.0%)	1 (14.3%)
Sixth or later	3 (1.81%)	0 (0.0%)	2 (66.7%)	1 (33.3%)	0 (0.0%)
Missing	10 (5.7%)
Third trimester nausea or vomiting
No nausea or vomiting	110 (62.5%)	31 (28.2%)	38 (34.6%)	24 (21.8%)	17 (15.5%)
Mild to moderate nausea or vomiting	65 (36.9%)	12 (18.5%)	26 (40.0%)	20 (30.8%)	7 (10.8%)
Missing	1 (0.6%)
Gestational age at birth (weeks)	39.0 (1.6)	39.1 (1.3)	39.2 (1.6)	38.8 (1.8)	38.9 (1.4)
Age (years)	28.6 (6.0)	27.8 (5.7)	30.0 (6.3)	27.8 (5.6)	29.1 (6.0)
Total GWG (kg)	10.5 (6.8)	13.1 (5.8)	11.8 (5.9)	9.1 (7.5)	4.6 (5.1)

GWG Distribution

Total GWG was lower in women who reported any nausea and vomiting in the third trimester (mean=8.7kg, SD=7.3) compared to women who reported no nausea and vomiting (mean=11.5kg, SD=6.2), t (173) =2.67, p=0.01. GWG was highest in nulliparous women (mean =12.8kg, SD=6.3). In contrast, for example, women who currently had more than five children gained significantly less weight (mean = 7.1, SD=8.1; F=−11.94, p<0.01). Older women also had lower total GWG (r=−0.21, p=0.005). Total GWG was highest amongst women with normal pre-pregnancy BMI and lowest in women with class 2/3 obesity (F=11.43, p<0.0001). Women with normal pre-pregnancy BMI gained an average of 13.1kg (SD=5.90), women with overweight BMI gained an average of 11.8kg (SD=5.90), women with class 1 obesity gained an average of 9.1kg (SD=7.52), and women with class 2/3 obesity gained an average of 4.6kg (SD=5.1). Distribution of GWG by pre-pregnancy BMI can be seen in Figure 2.

Figure 2.

Total Gestational Weight Gain (GWG) by Pre-Pregnancy BMI

Violin plot illustrating mean, standard deviation, and distribution of total GWG by pre-pregnancy BMI category

Note. Dots are mean total GWG per pre-pregnancy BMI category; lines are standard deviation per pre-pregnancy BMI category

AUCg and Diurnal Cortisol Levels by Pre-Pregnancy BMI

The average AUCg was 119.2 nmol/L (SD= 68.8) and ranged from 13.3 nmol/L to 424.9 nmol/L. AUCg distribution was right skewed, with the median value of 103.02 nmol/L and interquartile range of 69.15nmol/L. This distribution was in line with another study that also analyzed cortisol AUCg using the MADRES dataset.³⁷ Although AUCg for participants with pre-pregnancy BMI that was normal (mean= 125.9, SD=70.5) or overweight (mean=130.2, SD=76.7) were higher than those with class 1 (mean=105.8, SD=54.3) or class 2/3 obesity (mean=102.1, SD=63.5), these differences did not reach statistical significance (F=1.76, p=0.16).

However, cortisol levels at awakening (F=2.54, p=0.04) and 30 minutes after awakening (F=2.65, p=0.04) differed by pre-pregnancy BMI. Tukey’s test for post-hoc analysis showed that that awakening cortisol levels in women with class 1 obesity (β=−4.05, p=0.02) or class 2/3 obesity (β=−4.90, p=0.02) were lower than that in women with normal pre-pregnancy BMI (Figure 3). Though women with overweight pre-pregnancy BMI also showed lower awakening cortisol levels, the difference between them and women with normal pre-pregnancy BMI was not statistically significant (β=−2.72, p=0.09). Compared to women with normal pre-pregnancy BMI, cortisol levels 30 min after awakening in women with class 2/3 obesity was lower (β=−5.95, p=0.02). Though the magnitude of the effect was similar and directionality was the same, the differences were not statistically significant in women with class 1 obesity (β=−3.90, p=0.07) or overweight pre-pregnancy BMI (β=−0.81, p=0.68) compared to normal weight women. No differences were found in the afternoon (F=0.92, p=0.43) or evening (F=0.22, p=0.88) cortisol levels by pre-pregnancy BMI. Differences in other diurnal cortisol patterns such as the cortisol awakening response (F=1.27, p=0.29) and diurnal cortisol slope (F=2.39, p=0.07) by pre-pregnancy BMI were also examined but were not statistically significant.

Figure 3.

Diurnal Cortisol Patterns by Pre-Pregnancy BMI

AUCg and GWG

The association of total GWG with 3rd trimester AUCg varied by pre-pregnancy BMI category (3 df interaction test p=0.04). Figure 4 shows modeled total GWG as a function of 3rd trimester AUCg for women in each pre-pregnancy class. On average, amongst women whose pre-pregnancy BMI was Class 1 obese, a one standard deviation (1 SD) increase in 3rd trimester AUCg was associated with an increase in total GWG of 2.04 kg (95% CI: 0.17, 4.26; p=0.07). As shown in Table 2, there was no evidence for associations in the other pre-pregnancy BMI categories or overall. The stratified analysis examining the AUCg and total GWG relationship for each pre-pregnancy BMI category can be found in Supplemental Table 2. Though none of the estimates for each of the models reached statistical significance due to the limited sample size in each BMI group, the difference in magnitude and direction of the estimate can be seen by pre-pregnancy BMI category. For mothers with normal pre-pregnancy BMI, a 1SD increase in 3rd trimester AUCg was associated with an increase in total GWG of −0.91 kg (95% CI: −2.73, 0.91; p=0.31). On the other hand, a 1SD increase in 3rd trimester AUCg was associated with an increase in total GWG of 1.22kg (95% CI: −1.34, 3.78; p=0.34) for mothers with class 1 obese pre-pregnancy BMI.

Figure 4.

Association of Third Trimester Area Under the Curve and Gestational Weight Gain by Pre-Pregnancy BMI

Note: Model controlled for real-time personal monitoring sub-study participation (ref: not sub-study participant), gestational age at study entry, recruitment site (ref: LAC+USC), gestational age at birth, ethnicity by birthplace (ref: Non-Hispanic), highest education level (ref: High School), age, and third trimester nausea or vomiting (ref: no nausea or vomiting).

Table 2.

Estimated Difference in Total Gestational Weight Gain Associated with a 1 Standard Deviation Increase (68.82 nmol/L) in 3rd Trimester Cortisol Secretion (AUCg)

Model	BMI Class	Estimate (95% CI)	P-value
Model 1a	Overall	−0.11 (−0.99, 0.78)	0.82
Model 2b	Normal	−0.88 (−2.56, 0.79)	0.30
	Overweight	−0.41 (−1.70, 0.89)	0.54
	Class 1 Obese	2.04 (−0.17, 4.26)	0.07
	Class 2/3 Obese	0.08 (−2.48, 2.64)	0.95

^aModel 1 examined overall association between third trimester cortisol secretion and gestational weight gain

^bModel 2 examined differences in association between third trimester cortisol secretion and gestational weight gain by pre-pregnancy BMI

All models controlled for real-time personal monitoring sub-study participation (ref: not sub-study participant), gestational age at study entry, recruitment site (ref: LAC+USC), gestational age at birth, ethnicity by birthplace (ref: Non-Hispanic), highest education level (ref: High School), age, and third trimester nausea or vomiting (ref: no nausea or vomiting).

Discussion

The findings from the present study help to elucidate the complex interrelations between HPA axis related physiology, weight gain, and obesity in low-income, ethnic minority pregnant women. Though we did not find an association between third trimester total cortisol secretion and GWG overall, we discovered significant differences in this association by pre-pregnancy BMI. Women with class 1 obesity before pregnancy showed a more positive association between third trimester cortisol and GWG compared to women with normal BMI status before pregnancy. This finding can be seen whether we include pre-pregnancy BMI as a multiplicative interaction term or stratify by pre-pregnancy BMI. These results suggest that obesity status before pregnancy may impact the association between cortisol levels during late pregnancy and GWG. Our findings on pre-pregnancy BMI, prenatal circadian cortisol, and GWG are supported by previous evidence on obesity and HPA axis functioning.

Studies have shown the relationship between glucocorticoids and weight gain, specifically from increased adiposity, may differ by levels of abdominal obesity.³⁸ Though we are unable to differentiate pre-pregnancy levels of abdominal obesity vs. general obesity in this data set, findings from this study suggest that the metabolic mechanism of the HPA axis during pregnancy may be altered in women with class 1 obesity compared to those with normal pre-pregnancy BMI. It should be noted that contrary to our hypothesis, the association between prenatal circadian cortisol and GWG was not statistically different between normal weight women and women with class 2/3 obesity. This may be due to the limited sample size of women with class 2/3 obesity (n=24) or because a larger proportion of MADRES participants with class 2/3 obesity (62.5%) gained less than the recommended weight compared to the national average (21.9% of class 2 obese and 31.8% of class 3 obese). Since women with class 2/3 obesity in this data set did not experience much GWG, this may have attenuated any potential relationships with prenatal cortisol. Therefore, further research is needed in a population with higher class 2/3 obesity prevalence that is more representative of the national average.³⁹

Furthermore, as part of our ancillary analysis, our study found cortisol levels during awakening and 30 minutes after awakening to be significantly lower in women with obesity. Though the findings were not statistically significant, we also saw total cortisol levels to be lower in women with obese pre-pregnancy BMI. Previous studies conducted in pregnant populations have also identified lower cortisol levels earlier in the day in people with overweight or obesity. ^23,24,40 Overall cortisol and other HPA-related hormones are also decreased in pregnant women with obesity.^40,41 Our results support these findings and demonstrates the dysregulation of prenatal circadian HPA axis functioning in women with obesity.

Our results may indicate two separate physiological mechanisms that underscore the role of HPA axis functioning in GWG and birth outcomes. First, our findings showed total cortisol levels to be positively associated with GWG in women with class 1 obesity. This could imply that the relationship between cortisol and metabolic processes that stimulate weight gain during pregnancy are stronger in women in class 1 obesity compared to normal weight women.⁴² This is supported by previous studies that have shown an increased responsiveness of the HPA axis to food intake and vasopressin in ways that promote weight gain in those with obesity compared to those without. ^43,44 Second, women with pre-pregnancy obesity show decreased cortisol levels at awakening and 30 minutes after awakening that significantly depart from the expected pattern that is seen in women with normal pre-pregnancy BMI. Since cortisol plays an essential role in fetal organ maturation and initiating labor,⁴⁵ dysregulation of the HPA axis functioning may be contributing to the increased risk of negative developmental outcomes and prenatal complications, such as preterm birth and macrosomia that are seen in pregnant women with obesity. ^46–49 Both of our findings point to the potential importance of the contributing role of diurnal cortisol towards perinatal and birth outcomes. Given the disproportionate prevalence of adverse birth outcomes and pregnancy-related morbidities in women from lower-income and of racial and ethnic minorities, our study supports the importance of further elucidating HPA axis functioning in health disparities population.

Though this study presents novel findings on obesity, HPA axis functioning, and GWG in underrepresented pregnant women, this study has several limitations. Salivary samples were collected only once during the pregnancy with four samples on a single day to minimize participant and staff burden. Though one-time sampling protocol is not uncommon,^50–54 repeated assessment of salivary cortisol across the pregnancy is recommended to increase reliability.⁵⁵ In addition, when assessing salivary cortisol in epidemiological research, the gold standard includes at least six samples a day for up to six concurrent collection days in order to increase reliability and validity.^32,56 Another limitation is our lack of data regarding visceral adiposity, which plays an important role in obesogenic mechanisms of glucocorticoids. We acknowledge that time and contamination of saliva collection were self-reported, and though this is a common methodology, it is more prone to error than objective assessment methods.⁵⁷ Furthermore, pre-pregnancy BMI is based on self-reported weight, and therefore, may be at risk of recall bias, even with the extra precaution taken to eliminate any participants with a +/− 10kg discrepancy between the self-reported pre-pregnancy weight and first measured prenatal weight. Lastly, given that circadian cortisol was measured during the third trimester, we cannot establish causality on whether GWG up to the third trimester is influencing circadian cortisol or prenatal circadian cortisol is affecting GWG. Therefore, these results should be interpreted with caution.

Despite efforts to curtail obesity and excessive GWG rates, these issues remain highly prevalent among underrepresented women who are low-income and/or of color. The disproportionately high rates of obesity and excessive GWG in these populations are likely, in part, contributing to the increased risk of adverse pregnancy outcomes in these communities.⁵⁸ Findings from this study suggest the importance of further elucidating psychobiological mechanisms related to obesity and metabolism to continue addressing health disparities in maternal and infant outcomes.